Explicit public toNormal method about hyperloglog HOT 2 CLOSED

First of all thanks so much for taking the time to plot this... this is great

1. so Sparse improves the estimation of low cardinalities...
2. once somethign is normalized it can't be switched back
3. what about adding a NewNoSparse function that creates a HyperLogLog sketch without sparse?

from hyperloglog.

That works as well. Would you recommend that we always call NewNoSparse since our data sizes are always greater than 100?

Finally, since we're calling MarshalBinary to store the data and then fetching it later using UnmarshalBinary, how could we have UnmarshalBinary not call New and call NewNoSparse? Is there a way we can signal to UnmarshalBinary that we've already called New and it's initialized with the right precision?

from hyperloglog.